45 lines
1.2 KiB
Python
45 lines
1.2 KiB
Python
"""
|
|
@provider BaseLLMProvider (swarm/providers/base.py)
|
|
@purpose Abstract provider contract & token usage telemetry.
|
|
"""
|
|
from abc import ABC, abstractmethod
|
|
from typing import Type, TypeVar, List, Dict, Tuple
|
|
from pydantic import BaseModel, Field
|
|
|
|
T = TypeVar("T", bound=BaseModel)
|
|
|
|
|
|
class TokenUsage(BaseModel):
|
|
prompt_tokens: int = 0
|
|
completion_tokens: int = 0
|
|
total_tokens: int = 0
|
|
latency_ms: int = 0
|
|
model_name: str = "mock"
|
|
estimated_cost_usd: float = 0.0
|
|
|
|
|
|
class BaseLLMProvider(ABC):
|
|
@abstractmethod
|
|
async def generate_structured(
|
|
self,
|
|
prompt: str,
|
|
schema: Type[T],
|
|
system_prompt: str = ""
|
|
) -> Tuple[T, TokenUsage]:
|
|
"""
|
|
Executes a prompt and guarantees a response parsed into the requested Pydantic schema.
|
|
Returns the parsed model and precise token telemetry.
|
|
"""
|
|
pass
|
|
|
|
@abstractmethod
|
|
async def chat(
|
|
self,
|
|
messages: List[Dict[str, str]],
|
|
system_prompt: str = ""
|
|
) -> Tuple[str, TokenUsage]:
|
|
"""
|
|
Executes a multi-turn chat completion (used for interactive persona interviews).
|
|
Returns the response string and token telemetry.
|
|
"""
|
|
pass
|